Visualization Research Of Traditional Chinese Medicine Based On Three Dimensional Fluorescence Spectrum Characteristics On Pattern Recognition

Traditional Chinese medicine pattern recognition technology has important research significance in the quality of traditional Chinese medicine and new drug development.Feature extraction is a key technology in the field of traditional Chinese medicine pattern recognition research,but the characteristics extracted by different algorithms have a great influence on the classification and recognition effect.Therefore,choosing the appropriate feature extraction algorithm is extremely important in the classification identification research.In recent years,three-dimensional fluorescence spectroscopy has been widely used due to its sensitivity and rapidity.Most of the molecules in traditional Chinese medicine have the ability to generate fluorescence.Therefore,this paper introduces three-dimensional fluorescence spectroscopy into the research of traditional Chinese medicine classification and identification.The three-dimensional fluorescence spectrum characteristics of traditional Chinese medicine are classified and identified from the perspective of the efficacy of traditional Chinese medicine and the properties of traditional Chinese medicine.Firstly,based on the basic theory of efficacy and the properties of traditional Chinese medicine,the three-dimensional fluorescence spectrum characteristics of traditional Chinese medicine are analyzed.Aiming at the noise problem in the fluorescence spectrum signal of traditional Chinese medicine,based on the research of fluorescence spectrum denoising method of traditional Chinese medicine,the ensemble empirical mode decomposition algorithm is proposed to denoise the spectral noise of traditional Chinese medicine.Taking ginseng as an example to reduce noise,and comparing this method with wavelet decomposition and reconstruction,empirical mode decomposition algorithm,verify the superiority of EEMD algorithm noise reduction effect.Secondly,two-dimensional clustering and three-dimensional clustering visualization analysis of the main fluorescence peak position and main fluorescence peak intensity of different concentrations of traditional Chinese medicine solution.The fluorescence spectrum of 24 traditional Chinese medicines with the concentration of 10 mg/ml are used as an example to analyze the clustering effect of the main fluorescence peak position and intensity of the fluorescence spectrum of traditional Chinese medicine.Aiming at the efficacy of traditional Chinese medicine,the fitting regression relationship between the concentration of each solution and the intensity of the main fluorescence peak is studied quantitatively.From the qualitative point of view,the locally linear embedding algorithm is applied to extract the spectral data of tonic traditional Chinese medicines after noise reduction.Random forests algorithm is used for classification and recognition,and the accuracy is up to 95%.To verify the superiority of LLE,the algorithm is compared to the principal components analysis algorithm.Finally,according to the properties of traditional Chinese medicine,LLE and PCA algorithms are used to extract the fluorescence spectrum characteristics of cold and warm traditional Chinese medicines.The RF classifier is used to construct different classification models,and the spectral characteristics are classified and identified.The classification accuracy of LLE-RF and PCA-RF models are 96.6% and 89.7% respectively.Comparing the model constructed by RF classifier with the model constructed by support vector machine,the results show that the classification accuracy of LLE-SVM and PCA-SVM models are 82.1% and 89.5% respectively,indicating that the classification effect of LLE-RF model is better than other classification models.